如何解决自动参数选择过程-ARIMA / SARIMAX
无论如何，有没有阻止auto_arima函数到达给出无限aic的组合的功能。我正在使用逐步搜索，无法使用max_order作为参数。这是我的代码，
from pmdarima.arima import ndiffs,nsdiffs
test_df = grouped_df.get_group(items[k])
X = test_df['Quantity'].values
train,test = X[0:len(X)-1],X[len(X)-1:]
try:
nsdiffs_D = nsdiffs(X,m=52)
ndiffs_d = ndiffs(X,alpha=0.05)
stepwise_fit = auto_arima(train,start_p=0,start_q=0,max_p=6,max_q=6,m=52,start_P=0,seasonal=True,alpha=0.05,d=ndiffs_d,D=nsdiffs_D,trace=True,n_jobs=-1,error_action='ignore',stepwise=True,suppress_warnings=True)
在参数搜索中，大多数给出无限AIC的组合大部分是花费太多时间的组合。我正在对2000多种产品进行此网格搜索，当自动Arima对其中许多产品花费太多时间时，它没有用。是否有任何有关如何加快此网格搜索过程的建议。这是一次网格搜索的输出。
ARIMA(0,1,0)(0,1)[52] intercept : AIC=1699.129,Time=2.52 sec
ARIMA(0,0)[52] intercept : AIC=1701.752,Time=0.02 sec
ARIMA(1,0)(1,0)[52] intercept : AIC=1666.654,Time=3.13 sec
ARIMA(0,1)(0,1)[52] intercept : AIC=inf,Time=3.19 sec
ARIMA(0,0)[52] : AIC=1699.752,0)[52] intercept : AIC=1665.705,Time=0.12 sec
ARIMA(1,1)[52] intercept : AIC=1666.647,Time=2.60 sec
ARIMA(1,1)[52] intercept : AIC=1668.647,Time=5.35 sec
ARIMA(2,0)[52] intercept : AIC=1647.638,Time=0.07 sec
ARIMA(2,0)[52] intercept : AIC=1647.968,Time=5.39 sec
ARIMA(2,1)[52] intercept : AIC=1647.993,Time=3.75 sec
ARIMA(2,1)[52] intercept : AIC=1649.968,Time=8.39 sec
ARIMA(3,0)[52] intercept : AIC=1634.944,Time=0.10 sec
ARIMA(3,0)[52] intercept : AIC=1636.693,Time=8.38 sec
ARIMA(3,1)[52] intercept : AIC=1636.708,Time=5.15 sec
ARIMA(3,Time=14.20 sec
ARIMA(4,0)[52] intercept : AIC=1633.910,Time=0.14 sec
ARIMA(4,0)[52] intercept : AIC=1635.745,Time=2.38 sec
ARIMA(4,1)[52] intercept : AIC=1635.761,Time=6.20 sec
ARIMA(4,Time=16.45 sec
ARIMA(5,0)[52] intercept : AIC=1630.284,Time=0.41 sec
ARIMA(5,0)[52] intercept : AIC=1631.765,Time=6.28 sec
ARIMA(5,1)[52] intercept : AIC=1631.805,Time=5.67 sec
ARIMA(5,Time=15.37 sec
ARIMA(6,0)[52] intercept : AIC=1630.874,Time=0.45 sec
ARIMA(5,0)[52] intercept : AIC=inf,Time=0.66 sec
ARIMA(4,Time=0.56 sec
ARIMA(6,Time=0.80 sec
ARIMA(5,0)[52] : AIC=1628.369,Time=0.20 sec
ARIMA(5,0)[52] : AIC=1629.844,Time=3.39 sec
ARIMA(5,1)[52] : AIC=1629.883,Time=3.79 sec
ARIMA(5,1)[52] : AIC=inf,Time=5.90 sec
ARIMA(4,0)[52] : AIC=1631.963,Time=0.08 sec
ARIMA(6,0)[52] : AIC=1628.991,Time=0.26 sec
ARIMA(5,0)[52] : AIC=1614.244,Time=0.47 sec
ARIMA(5,1)(1,0)[52] : AIC=1614.307,Time=6.95 sec
ARIMA(5,1)[52] : AIC=1614.428,Time=4.40 sec
ARIMA(5,Time=7.27 sec
ARIMA(4,0)[52] : AIC=1613.564,Time=0.23 sec
ARIMA(4,0)[52] : AIC=1614.206,Time=3.28 sec
ARIMA(4,1)[52] : AIC=1614.355,Time=3.48 sec
ARIMA(4,Time=8.28 sec
ARIMA(3,0)[52] : AIC=1611.646,Time=0.29 sec
ARIMA(3,0)[52] : AIC=1612.206,Time=3.24 sec
ARIMA(3,1)[52] : AIC=1612.356,Time=2.45 sec
ARIMA(3,Time=5.55 sec
ARIMA(2,0)[52] : AIC=1610.019,Time=0.12 sec
ARIMA(2,0)[52] : AIC=1610.438,Time=3.43 sec
ARIMA(2,1)[52] : AIC=1610.594,Time=2.02 sec
ARIMA(2,Time=4.26 sec
ARIMA(1,0)[52] : AIC=1609.541,Time=0.07 sec
ARIMA(1,0)[52] : AIC=1610.207,Time=1.71 sec
ARIMA(1,1)[52] : AIC=1610.309,Time=1.88 sec
ARIMA(1,1)[52] : AIC=1612.024,Time=7.56 sec
ARIMA(0,0)[52] : AIC=1608.209,Time=0.05 sec
ARIMA(0,0)[52] : AIC=1608.533,Time=0.97 sec
ARIMA(0,1)[52] : AIC=1608.656,Time=1.01 sec
ARIMA(0,1)[52] : AIC=1610.357,Time=2.37 sec
ARIMA(0,2)(0,0)[52] : AIC=1609.353,Time=0.11 sec
ARIMA(1,0)[52] : AIC=1663.708,0)[52] : AIC=1611.960,Time=0.17 sec
ARIMA(0,Time=0.12 sec
Best model: ARIMA(0,0)[52]
Total fit time: 203.303 seconds
版权声明：本文内容由互联网用户自发贡献，该文观点与技术仅代表作者本人。本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。如发现本站有涉嫌侵权/违法违规的内容， 请发送邮件至 dio@foxmail.com 举报，一经查实，本站将立刻删除。